The computational cost of quasi-P wave extrapolation depends on the complexity of the medium, and specifically the anisotropy. Our effective-model method splits the anisotropic dispersion relation into an isotropic background and a correction factor to handle this dependency. The correction term depends on the slope (measured using the gradient) of current wavefields and the anisotropy. As a result, the computational cost is independent of the nature of anisotropy, which makes the extrapolation efficient. A dynamic implementation of this approach decomposes the original pseudo-differential operator into a Laplacian, handled using the low-rank approximation of the spectral operator, plus an angular dependent correction factor applied in the space domain to correct for anisotropy. We analyze the role played by the correction factor and propose a new spherical decomposition of the dispersion relation. The proposed method provides accurate wavefields in phase and more balanced amplitudes than a previous spherical decomposition. Also, it is free of SV-wave artifacts. Applications to a simple homogeneous transverse isotropic medium with a vertical symmetry axis (VTI) and a modified Hess VTI model demonstrate the effectiveness of the approach. The Reverse Time Migration (RTM) applied to a modified BP VTI model reveals that the anisotropic migration using the proposed modeling engine performs better than an isotropic migration.
Bibliographical noteKAUST Repository Item: Exported on 2020-10-01
Acknowledgements: We thank Nabil Masmoudi and Hui Wang for their helpful discussions, and Rene-Edouard Plessix, Samuel Gray and one anonymous reviewer for helpful reviews. We thank KAUST for its support and specifically the seismic wave analysis group members for their valuable insights. The published results are reproducible from the open source software Madagascar (www.ahay.org). The research was partly funded by the National Natural Science Foundation of China (Grant Nos. 41730425).